The use of artificial intelligence (AI) is becoming increasingly popular and in demand in the modern world. Companies are seeking to implement AI technologies to increase the efficiency of their business, optimise processes,as well as improve the quality of their products. However, the implementation and maintenance of AI solutions requires special tools and platforms capable of ensuring reliability, scalability and manageability of projects.

Yaroslav Starukhin, Senior Data Scientist, an expert in AI product development and its implementation, shares his experience in building an MLOps platform for the oil industry.

Yaroslav, how did you get into IT? Have you always been interested in technology? 

Already at school, I was interested in mathematics and computer sciencу—these subjects came easily to me, and I could cope with complex mathematical problems and programming. My passion for these fields made me clearly define my future career path, and I entered the field of applied mathematics and computer science to get a profound education.

I received my first education at the Plekhanov Russian University of Economics, where I studied mathematical methods in economics. This gave me a basic knowledge of various areas of mathematics, software engineering, and data analysis, which formed the basis of my future professional activity in the field of analytics and information technology.

I then continued my studies at the Yandex School of Data Analytics, a prestigious institution in the field of data science and artificial intelligence. Here I deepened my knowledge of machine learning by taking advanced courses in programming, software architecture, Bayesian statistics, language technologies and other modern tools that are increasingly in demand in the industry.

I took my first steps into the industry during my internship in 2016 while I was still a student. Since then, I have gained a lot of valuable experience in the IT field. Currently, all of my professional activities are closely connected with my education and hobbies, and it is a pleasure to be able to apply my knowledge and skills to large-scale projects.

Tell us more about your projects: where were you working precisely? Which were the challenges you faced? 

Let's order: I worked for one of the major retailers in Eastern Europe, which was a federal-scale grocer with a high level of service and substantial investment. The customers are a premium segment of clientele with high brand loyalty. Chain shops are represented in all cities of the country, and the distinctive feature is personal service, which includes individual promotions and special offers in accordance with the loyalty programme. Loyalty cards allow the accumulation of purchase history for each customer, which further enables the shop to create these personalised promotional campaigns and improve communication with each customer.

The aim of this project is to improve the customer experience and enhance the shop's financial metrics by shifting from mass promos to targeted ones, which usually results in increased conversions and revenue growth for the company. This kind of personalisation helps to stay competitive and stand out among other industry players. Therefore, companies that neglect such projects are at risk of losing market share.

Upon joining the team, I was given the challenging task of organising the analytical work in this project. I designed the complete process that the team goes through from idea to large-scale campaign launch. The process involved interaction between different cross-functional team specialists and a lot of data analytics: in order to launch such a personalised promotional campaign, you need to go through hypothesis generation, simulate the campaign on historical data, run an AV experiment for a small group of end users, evaluate the effectiveness and scalability of the campaign, decide on a full-scale launch, and automate such launches for the whole network taking into account various constraints such as frequency of communication. I was engaged in implementing and testing the first versions of such promotional campaigns to fine-tune the described analytical processes: I myself developed machine learning models for predicting responses to promos, segmenting customer groups, detecting churn/ infrequent users, building AB tests to evaluate campaign results and generated the automatic pipelines that helped automate these processes.

Once the test turned out to be successful, the challenge was to scale the project internally by bringing in new talent, interviewing and training them on the workflow. My role was to train and educate the team and expand the product to all shop branches for continued success in the market.

I also worked on a project that was similar, but on a larger scale, for a major Russian retailer. The main difference was that in the CIS network, there were about seven million loyal users with registered cards, whereas in Russia, their number reached twenty-five million. The scale gap is very substantial. Another problem was the difference in geography: Russia has a much larger area, different cities, and it resulted in disparities in operations, shops, logistics, and assortment. My task was to help the company to effectively set up the whole process of work, as well as to improve the personal communications of shops with customers. We eventually built processes to evaluate their financial performance. As part of the work, the project launched many personal promos, and it is necessary to test various hypotheses to understand which of them turned out to be correct, where the company was able to achieve results, and where the company worked at zero. To address these questions, I developed a platform for AV testing and processes related to evaluating the financial effectiveness of promos.

From the client's side, our team identified different case studies: expanding the product basket, increasing the frequency of visits, reducing churn, and others. The tool I developed allowed me to test hypotheses (pilot launches of advertising campaigns for small audiences)—I was able to conduct AB experiments, given the random nature of the results. That's why we need to understand whether the same positive effect will be achieved when the campaign is replicated for a larger audience.

Yaroslav, you have extensive experience in modern retail. What other sectors have you worked in? 

There is a large copper and ore processing holding in Kazakhstan. Its plants are of federal importance and operate with huge budgets, which is important not only for the company but also for the country itself. The plants operate in different shifts, each with its own management experience. However, when I arrived, there was no understanding of how data could be used to combine this experience from different shifts to optimise ore processing.

To address this, a digital advisor was developed, a web page located at the head office. Plant operators can use the Digital Advisor to receive recommendations on how to optimise the management of the units—for example, how to change the grinding and reagent levels to increase the recovery of valuable metal. These recommendations are generated by the AI system and require further acceptance or rejection, resulting in changes to plant operations.

This project involved a significant transformation, as the factory's employees were used to a certain way of working and distrustful of AI applications. In order to change the approach to working in and managing the factory, the aim was to start making data-driven decisions and extract useful insights through similar digital advisors.

The project also included updating KPIs, conducting training, and educating employees on new management techniques. Ultimately, the company successfully implemented the transformation at one of the factories and then decided to extend the project to another. 

Another industry in which I was fortunate enough to gain experience in applying data analytics was the automotive industry. A well-known automotive company with Japanese roots manufactures its cars in various countries around the world, has many factories and outlets, and makes its products available in each country. However, because of the company's long history, its internal processes are not always in line with best practices, which hampers decision-making. The company's business processes are shaped by Japanese philosophy, which means that any changes take time and go through voting and discussion processes, which slows down innovation. As a result, the company is not always agile in planning the production and distribution of its vehicles.

The project I worked on involved the development of a planning system that produced optimal assembly and distribution plans for machines by country. This planning system made it possible to optimise production processes and distribution plans, taking into account the specifics of each country and market requirements. The car production and distribution process has many aspects, such as storage constraints for the cars produced, different demands in the final regions, capacity constraints on the assembly lines, different levels of profitability in different markets, and other business aspects that need to be taken into account.

The development of this system has been a complex task, involving the creation of various components and modules, as well as integration with the work of staff at all levels. The changes in business processes associated with the implementation of the new system represented a significant challenge for a company that had long operated according to established patterns.

My role was to develop a mathematical algorithm for planning based on business requirements from the planning and sales departments. It was also important to create transparency in the operation of the system, explaining how it worked and its limitations, so that people could better understand the decisions being made and have confidence in the system. 

It took considerable effort and time to successfully integrate and use the new system, but ultimately the project was successful and the company became more efficient and adaptable to changing market conditions due to the shortened planning cycles.

What are you working on now?

I am building an MLOps platform for the largest oil company outside the Arab world. The project is part of a major digital transformation, which aims to align horizontal and vertical products in the company. A horizontal product is functional expertise and everything that goes with it, while a vertical product is a product for a specific use case. For example, a vertical product is a digital advisor that provides a recommendation on how to drill a well every five minutes, and a horizontal product is an MLOps platform that allows you to scale and automate these types of products.

The machine learning patterns that underpin such AI products require constant monitoring, preferably without human input. This requires timely and accurate reporting on the quality of the data and the functioning of the models, as well as verification of the results achieved with such products, and so on. Another feature of this project is that I am creating a system not just for one AI product, but to set standards for all of the company's future products.

Why is this project so important?

The importance of this project is critical to the client and their business. Currently, there is no such combination of team, tools and processes that can enable a company to quickly, efficiently and sustainably develop dozens of AI products using hundreds of machine learning models. It all depends on specific data science experts and machine learning engineers. Once the "horizontality" of MLOps is in place, the time required to develop an AI product is reduced from around 5–6 months to 6–7 weeks.

What should be the outcome of the completed project? 

The outcome of the project will be to generate immediate revenue for the company by launching the products and to ensure that the effectiveness of the products is maintained over time after six months of operation without constant monitoring by the team. Having a system in place to ensure the sustainability and long-term performance of AI products is key to a successful business in today's world.